Update on Week 3–4 Severe Weather Project
9 February 2018
Week 3–4 Severe Weather Project
Forecasts for weeks 1–2 are just a stepping stone.
Work is under progress for weeks 3–4 using the CFSv2 45-day hindcasts and real-time forecasts.
Methodology
Hybrid dynamical–statistical forecast model
Using the Supercell Composite Parameter (SCP) from the GEFS forecasts as a predictor
Week 1–2 SW forecast based on the statistical relationship between model SCP (GEFS hindcast) and OBS LSR (Local Storm Report), including hail, tornado, and damaging wind
Forecast improvements:
• 5o×5o area-average anomalies
• Adjustment for the amplitude of forecasted anomalies
Forecast skill (AC) assessed with cross-validations over the 1996–2012 GEFS hindcast period
SCP (Supercell Composite Parameter)
SCP = (CAPE/1000 J kg1)×(SRH/50 m2 s2)×(BWD/20 m s1)
CAPE: convective available potential energy
SRH: storm-relative helicity
BWD: bulk wind difference
Week 1 GEFS Skill for SCP
AC: GEFS vs. CFSR 1996–2012
GEFS Hindcasts
16-day forecast
1996–2012 (17yrs)
4 days apart
5 members
0.5o×0.5o resolution
Week-2 skill is much lower than the Week-1.
Cross-validation over MAM 1996–2012 5o×5o area average
Hybrid Model Forecast Skill for LSR3 Anomaly Correlation
Relationship between GEFS SCP and OBS LSR3
LSR3: Hail + Tornado + Wind
Tornado Outbreaks May 24–30, 2011
Week-1 Forecast LSR3 Week-2 Forecast LSR3
Week-1 Forecast (Adjusted) Week-2 Forecast (Adjusted) 5o×5o 0.5o×0.5o
OBS Weekly LSR3 OBS Weekly LSR3
Forecast = AnomalyFCST
+ ClimatologyOBS
Adjustment =
AnomalyFCST × Ratio + ClimatologyOBS Ratio = SDOBS/SDFCST
GEFS Real-time Forecast for SCP
Weekly SCP is the 7-day average of daily mean SCP from 12Z to 12Z.
Day 1 12Z
18Z
Day 0 00Z
06Z
Day 0 12Z
20 x 4 = 80 members
Week 1 Week 2
GEFS 80-member Mean Forecast
04Feb 12Z 18Z 05Feb 00Z 06Z
Week 1
Week 2
SCP 5o×5
o SPC Convective Outlook
Day 1
Day 2
Day 3
Week-1 Forecast LSR3
Week-1 Forecast Category and Probability
Week 1:
5–11 FEB
2018 80-member
ensemble
mean forecast
Based on the distribution of the 80-member forecasts
Probabilistic Outlook
Distribution of LSR3 hindcasts (or observations) determines the thresholds for above-normal, near-normal, and below- normal categories, each 33%.
The distribution of individual 80-member forecasts in the three categories provides the probabilistic forecast for LSR3.
In few months, we can start providing real-time week 1–2 severe weather forecasts based on GEFS to OPB.